diff --git a/frontend/src/components/protocol/AddNodePanel.tsx b/frontend/src/components/protocol/AddNodePanel.tsx
index 4a95bc4..3cc71ee 100644
--- a/frontend/src/components/protocol/AddNodePanel.tsx
+++ b/frontend/src/components/protocol/AddNodePanel.tsx
@@ -1,5 +1,5 @@
 import { useState } from 'react'
-import { ArrowRight, Atom, BookMarked, Bot, Braces, BrainCircuit, Cloud, Code2, Cog, Database, FileText, HardDrive, Repeat2, Route, ScrollText, ShieldCheck, Server, Sparkles, X } from 'lucide-react'
+import { ArrowRight, Atom, BookMarked, Bot, Braces, BrainCircuit, Cloud, Code2, Cog, Database, FileText, HardDrive, Repeat2, Route, ScrollText, ShieldCheck, Server, Sparkles, User, X } from 'lucide-react'
 import { Button } from '@/components/ui/button'
 import { Input } from '@/components/ui/input'
 import { DATASET_BROWSE } from './datasetCatalog'
@@ -107,6 +107,13 @@ const NODE_CATALOG = [
     description: 'Not yet functional -- declares intent for a future phase',
     icon: BrainCircuit,
   },
+  {
+    type: 'persona',
+    category: 'knowledge',
+    label: 'Persona',
+    description: 'OCEAN personality traits for experimental manipulation of agent behavior',
+    icon: User,
+  },
   // Not a node type either -- picking this opens the dataset browser
   // (DatasetBrowserPanel), same reasoning as MCP Servers above and Skills
   // below: the
diff --git a/frontend/src/components/protocol/DesignTab.tsx b/frontend/src/components/protocol/DesignTab.tsx
index 369919e..b7d56d6 100644
--- a/frontend/src/components/protocol/DesignTab.tsx
+++ b/frontend/src/components/protocol/DesignTab.tsx
@@ -504,6 +504,8 @@ export function DesignTab({
       // too -- not just the cell rows.
       queryClient.invalidateQueries({ queryKey: ['experiments', experiment.id, 'design-revisions'] })
       queryClient.invalidateQueries({ queryKey: ['experiments', experiment.id, 'design-impact'] })
+      // Invalidate runs list which shows cell labels
+      queryClient.invalidateQueries({ queryKey: ['protocol-runs'] })
     },
   })
 
diff --git a/frontend/src/components/protocol/FactorEditorDialog.tsx b/frontend/src/components/protocol/FactorEditorDialog.tsx
index 35ae361..dce7e09 100644
--- a/frontend/src/components/protocol/FactorEditorDialog.tsx
+++ b/frontend/src/components/protocol/FactorEditorDialog.tsx
@@ -39,7 +39,9 @@ import { PythonCodeEditor } from './PythonCodeEditor'
 import { useDialogAutosave } from './useDialogAutosave'
 import { useDialogDraft } from './useDialogDraft'
 import type { DesignFactor } from '@/types/experiments'
+import { Library, Upload } from 'lucide-react'
 import { DatasetFactorEditor } from './DatasetFactorEditor'
+import { PERSONA_LIBRARY } from './personaLibrary'
 import { KnowledgeFactorEditor } from './KnowledgeFactorEditor'
 import { ScriptFactorEditor } from './ScriptFactorEditor'
 import { ToolFactorEditor } from './ToolFactorEditor'
@@ -419,6 +421,195 @@ function DatasetConfigLevelRow({
   )
 }
 
+function PersonaLevelRow({ value, onChange }: { value: StructuredLevel; onChange: (next: StructuredLevel) => void }) {
+  const [selectedPersonas, setSelectedPersonas] = useState<string[]>(() => {
+    const ids = value.persona_ids as string[] | undefined
+    if (ids && Array.isArray(ids) && ids.length > 0) return ids
+    const id = value.persona_id as string | null
+    return id ? [id] : ['']
+  })
+
+  // Sync state when value changes (e.g., when loading an existing factor for editing)
+  useEffect(() => {
+    const ids = value.persona_ids as string[] | undefined
+    if (ids && Array.isArray(ids) && ids.length > 0) {
+      setSelectedPersonas(ids)
+    } else {
+      const id = value.persona_id as string | null
+      setSelectedPersonas(id ? [id] : [''])
+    }
+  }, [value.persona_ids, value.persona_id])
+
+  function patch(patch: StructuredLevel) {
+    onChange({ ...value, ...patch })
+  }
+
+  function updatePersona(index: number, newId: string | null) {
+    const newSelected = [...selectedPersonas]
+    newSelected[index] = newId || ''
+    const filtered = newSelected.filter(id => id !== '')
+    setSelectedPersonas(filtered.length > 0 ? filtered : [''])
+
+    const combinedText = filtered
+      .map(id => PERSONA_LIBRARY.find(p => p.id === id)?.prompt ?? '')
+      .filter(Boolean)
+      .join('\n\n')
+
+    const allLabels = filtered
+      .map(id => PERSONA_LIBRARY.find(p => p.id === id)?.label)
+      .filter(Boolean)
+      .join(' + ')
+
+    patch({
+      persona_text: combinedText,
+      persona_id: filtered[0] || null,
+      persona_ids: filtered,
+      persona_label: allLabels || null,
+      mode: 'library',
+      persona_file: null
+    })
+  }
+
+  function addPersona() {
+    if (selectedPersonas.length < 5) {
+      setSelectedPersonas([...selectedPersonas, ''])
+    }
+  }
+
+  function removePersona(index: number) {
+    const newSelected = selectedPersonas.filter((_, i) => i !== index)
+    setSelectedPersonas(newSelected.length > 0 ? newSelected : [''])
+
+    const combinedText = newSelected
+      .map(id => PERSONA_LIBRARY.find(p => p.id === id)?.prompt ?? '')
+      .filter(Boolean)
+      .join('\n\n')
+
+    const allLabels = newSelected
+      .map(id => PERSONA_LIBRARY.find(p => p.id === id)?.label)
+      .filter(Boolean)
+      .join(' + ')
+
+    patch({
+      persona_text: combinedText,
+      persona_id: newSelected[0] || null,
+      persona_ids: newSelected,
+      persona_label: allLabels || null,
+      mode: newSelected.length > 0 ? 'library' : 'custom',
+      persona_file: null
+    })
+  }
+
+  async function handleFileUpload(e: React.ChangeEvent<HTMLInputElement>) {
+    const file = e.target.files?.[0]
+    if (!file) return
+    const text = await file.text()
+    patch({
+      persona_text: text,
+      persona_file: file.name,
+      mode: 'file',
+      persona_id: null,
+      persona_ids: [],
+      persona_label: null
+    })
+    setSelectedPersonas([''])
+  }
+
+  return (
+    <div className="space-y-2 rounded-lg border p-2">
+      <div className="space-y-1.5">
+        <Label className="text-xs">Choose from library</Label>
+        <div className="space-y-1.5">
+          {selectedPersonas.map((personaId, index) => (
+            <div key={index} className="flex gap-1.5">
+              <Select
+                value={personaId || '__none__'}
+                onValueChange={(newId) => {
+                  if (newId === '__none__') return
+                  updatePersona(index, newId)
+                }}
+              >
+                <SelectTrigger className="h-8 flex-1">
+                  <SelectValue>
+                    {() => PERSONA_LIBRARY.find(p => p.id === personaId)?.label ?? 'Select a persona...'}
+                  </SelectValue>
+                </SelectTrigger>
+                <SelectContent>
+                  <SelectItem value="__none__" disabled>
+                    Select a persona...
+                  </SelectItem>
+                  {PERSONA_LIBRARY.map(p => (
+                    <SelectItem key={p.id} value={p.id}>
+                      <div className="flex items-center gap-2">
+                        <Library className="size-4" />
+                        {p.label}
+                      </div>
+                    </SelectItem>
+                  ))}
+                </SelectContent>
+              </Select>
+              {selectedPersonas.length > 1 && (
+                <Button
+                  variant="ghost"
+                  size="sm"
+                  type="button"
+                  className="h-8 w-8 p-0"
+                  onClick={() => removePersona(index)}
+                >
+                  <X className="size-3.5" />
+                </Button>
+              )}
+            </div>
+          ))}
+        </div>
+        {selectedPersonas.length < 5 && (
+          <Button
+            variant="outline"
+            size="sm"
+            type="button"
+            className="h-7 w-full text-xs"
+            onClick={addPersona}
+          >
+            <Plus className="size-3 mr-1" />
+            Add another (max 5)
+          </Button>
+        )}
+      </div>
+      <div className="space-y-1">
+        <Label className="text-xs">Upload .md file</Label>
+        <div className="flex gap-2">
+          <Button
+            variant="outline"
+            size="sm"
+            type="button"
+            className="h-8 w-full text-xs"
+            onClick={() => document.getElementById(`persona-upload-${String(value.persona_id || 'custom')}`)?.click()}
+          >
+            <Upload className="size-3 mr-1.5" />
+            {(value.persona_file as string) || 'Choose .md file'}
+          </Button>
+          <input
+            id={`persona-upload-${String(value.persona_id || 'custom')}`}
+            type="file"
+            accept=".md"
+            className="hidden"
+            onChange={handleFileUpload}
+          />
+        </div>
+      </div>
+      <div className="space-y-1">
+        <Label className="text-xs">Persona text</Label>
+        <Textarea
+          value={(value.persona_text as string) ?? ''}
+          onChange={(e) => patch({ persona_text: e.target.value, mode: 'custom', persona_id: null, persona_label: null, persona_file: null })}
+          placeholder="You are extremely organized, very responsible..."
+          className="h-20 font-mono text-xs"
+        />
+      </div>
+    </div>
+  )
+}
+
 // One row of a "tool_names" factor's levels -- an allow-list of BARE tool
 // names for one MCP node, i.e. exactly what McpToolNodeInspector's own "Tools
 // allowed" toggle list writes into config.tool_names (protocol_execution.py's
@@ -665,7 +856,18 @@ function StandardFactorEditorDialog({
           ? [false, true]
           : selectedIndexes.map((index) => parseLevelValue((draftLevels as string[])[index], draftLevelType))
     const defaults = defaultFactorLevelLabels(draftName, parsedLevels.length, draftLevelType)
-    const labels = selectedIndexes.map((index, outputIndex) => draftLevelLabels[index]?.trim() || defaults[outputIndex])
+    const labels =
+      draftLevelType === 'persona_text'
+        ? draftLevels.map((level, index) => {
+            if (typeof level === 'object' && level !== null && 'persona_label' in level) {
+              const personaLabel = (level as Record<string, unknown>).persona_label
+              if (typeof personaLabel === 'string' && personaLabel.trim()) {
+                return personaLabel.trim()
+              }
+            }
+            return defaults[index]
+          })
+        : selectedIndexes.map((index, outputIndex) => draftLevelLabels[index]?.trim() || defaults[outputIndex])
     return { factor: { name: draftName, levels: parsedLevels, level_labels: labels, level_type: draftLevelType } satisfies DesignFactor, field }
   }
 
@@ -894,6 +1096,11 @@ function StandardFactorEditorDialog({
                               }
                               onChange={(next) => setLevels((ls) => ls.map((l, j) => (j === i ? next : l)))}
                             />
+                          ) : levelType === 'persona_text' ? (
+                            <PersonaLevelRow
+                              value={level as StructuredLevel}
+                              onChange={(next) => setLevels((ls) => ls.map((l, j) => (j === i ? next : l)))}
+                            />
                           ) : (
                             <ScriptConfigLevelRow
                               value={level as StructuredLevel}
diff --git a/frontend/src/components/protocol/PersonaNodeInspector.tsx b/frontend/src/components/protocol/PersonaNodeInspector.tsx
new file mode 100644
index 0000000..d1c404b
--- /dev/null
+++ b/frontend/src/components/protocol/PersonaNodeInspector.tsx
@@ -0,0 +1,296 @@
+import { useState } from 'react'
+import { nodeAccent } from '@/lib/nodeAccent'
+import { User, Upload, Library, Plus, X } from 'lucide-react'
+import { Label } from '@/components/ui/label'
+import { Textarea } from '@/components/ui/textarea'
+import { Button } from '@/components/ui/button'
+import { Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from '@/components/ui/select'
+import { EditableNodeTitle } from './EditableNodeTitle'
+import { FactorBindableField } from './FactorBindableField'
+import { NodeInspectorDialog } from './NodeInspectorDialog'
+import { PERSONA_LIBRARY } from './personaLibrary'
+import type { ProtocolNode, PersonaNodeData } from '@/types/protocols'
+
+const ACCENT = nodeAccent('persona')
+
+
+export function PersonaNodeInspector({
+  node,
+  experimentId,
+  factorNodeLabel,
+  onChange,
+  onDelete,
+  onClose,
+}: {
+  node: (ProtocolNode & { data: PersonaNodeData }) | null
+  experimentId: string | null
+  factorNodeLabel: string
+  onChange: (nodeId: string, data: PersonaNodeData) => void
+  onDelete: (nodeId: string) => void
+  onClose: () => void
+}) {
+  // Track multiple selected personas for combination (can include empty strings for unfilled slots)
+  const [selectedPersonas, setSelectedPersonas] = useState<string[]>(() => {
+    if (!node) return []
+    const config = node.data.config
+    // Initialize from persona_ids array if present, otherwise fall back to single persona_id
+    if (config.persona_ids && Array.isArray(config.persona_ids) && config.persona_ids.length > 0) {
+      return config.persona_ids as string[]
+    }
+    return config.persona_id ? [config.persona_id] : []
+  })
+
+  if (!node) return null
+  const nodeId = node.id
+  const data = node.data
+  const config = data.config
+  const bindings = data.factor_bindings ?? {}
+
+  function patchConfig(patch: Partial<PersonaNodeData['config']>) {
+    onChange(nodeId, { ...data, config: { ...config, ...patch } })
+  }
+
+  function bindFactor(fieldPath: string, factorName: string) {
+    onChange(nodeId, { ...data, factor_bindings: { ...bindings, [fieldPath]: factorName } })
+  }
+
+  function unbindFactor(fieldPath: string) {
+    const next = { ...bindings }
+    delete next[fieldPath]
+    onChange(nodeId, { ...data, factor_bindings: next })
+  }
+
+  async function handleFileUpload(e: React.ChangeEvent<HTMLInputElement>) {
+    const file = e.target.files?.[0]
+    if (!file) return
+    const text = await file.text()
+    patchConfig({
+      persona_text: text,
+      persona_file: file.name,
+      mode: 'file'
+    })
+  }
+
+  return (
+    <NodeInspectorDialog
+      open={true}
+      onOpenChange={onClose}
+      accent={ACCENT}
+      title={
+        <>
+          <User className="size-5" style={{ color: ACCENT }} />
+          <EditableNodeTitle label={data.label} placeholder="Persona" onCommit={(label) => onChange(nodeId, { ...data, label })} />
+        </>
+      }
+      onDelete={() => onDelete(nodeId)}
+      onClose={onClose}
+    >
+      <div className="space-y-6">
+
+        {/* Choose from library - multi-select to combine personas */}
+        <div className="space-y-2">
+          <div className="flex items-center justify-between">
+            <Label>Choose from library</Label>
+            {selectedPersonas.length > 0 && selectedPersonas.length < 5 && (
+              <Button
+                variant="ghost"
+                size="sm"
+                onClick={() => setSelectedPersonas([...selectedPersonas, ''])}
+                className="h-7 text-xs"
+              >
+                <Plus className="size-3 mr-1" />
+                Add another
+              </Button>
+            )}
+          </div>
+
+          {selectedPersonas.length === 0 ? (
+            <Select
+              value=""
+              onValueChange={(personaId) => {
+                const persona = PERSONA_LIBRARY.find(p => p.id === personaId)
+                if (persona && personaId) {
+                  setSelectedPersonas([personaId])
+                  patchConfig({
+                    persona_text: persona.prompt || '',
+                    persona_id: persona.id || null,
+                    persona_label: persona.label || null,
+                    mode: 'library'
+                  })
+                }
+              }}
+            >
+              <SelectTrigger>
+                <SelectValue placeholder="Select a validated persona..." />
+              </SelectTrigger>
+              <SelectContent>
+                {PERSONA_LIBRARY.map(p => (
+                  <SelectItem key={p.id} value={p.id}>
+                    <div className="flex items-center gap-2">
+                      <Library className="size-4" />
+                      {p.label}
+                    </div>
+                  </SelectItem>
+                ))}
+              </SelectContent>
+            </Select>
+          ) : (
+            <div className="space-y-2">
+              {selectedPersonas.map((personaId, index) => (
+                <div key={index} className="flex items-center gap-2">
+                  <Select
+                    value={personaId || ''}
+                    onValueChange={(newId) => {
+                      const newSelected = [...selectedPersonas]
+                      newSelected[index] = newId || ''
+                      const filtered = newSelected.filter(id => id)
+                      setSelectedPersonas(newSelected)
+
+                      // Combine all selected persona texts
+                      const combinedText = filtered
+                        .map(id => PERSONA_LIBRARY.find(p => p.id === id)?.prompt ?? '')
+                        .filter(text => text)
+                        .join('\n\n')
+
+                      const labels = filtered
+                        .map(id => PERSONA_LIBRARY.find(p => p.id === id)?.label)
+                        .filter(Boolean)
+                        .join(' + ')
+
+                      patchConfig({
+                        persona_text: combinedText,
+                        persona_id: filtered[0] || null,
+                        persona_ids: filtered, // Store all IDs for multi-select restoration
+                        persona_label: labels || null,
+                        mode: 'library'
+                      })
+                    }}
+                  >
+                    <SelectTrigger className="flex-1">
+                      <SelectValue placeholder={`Select persona ${index + 1}...`} />
+                    </SelectTrigger>
+                    <SelectContent>
+                      {PERSONA_LIBRARY.map(p => (
+                        <SelectItem key={p.id} value={p.id}>
+                          <div className="flex items-center gap-2">
+                            <Library className="size-4" />
+                            {p.label}
+                          </div>
+                        </SelectItem>
+                      ))}
+                    </SelectContent>
+                  </Select>
+                  <Button
+                    variant="ghost"
+                    size="icon-sm"
+                    onClick={() => {
+                      const newSelected = selectedPersonas.filter((_, i) => i !== index)
+                      setSelectedPersonas(newSelected)
+
+                      if (newSelected.length === 0) {
+                        patchConfig({
+                          persona_text: '',
+                          persona_id: null,
+                          persona_ids: [],
+                          persona_label: null,
+                          mode: 'text'
+                        })
+                      } else {
+                        // Recombine remaining personas
+                        const filtered = newSelected.filter(id => id)
+                        const combinedText = filtered
+                          .map(id => PERSONA_LIBRARY.find(p => p.id === id)?.prompt ?? '')
+                          .filter(text => text)
+                          .join('\n\n')
+
+                        const labels = filtered
+                          .map(id => PERSONA_LIBRARY.find(p => p.id === id)?.label)
+                          .filter(Boolean)
+                          .join(' + ')
+
+                        patchConfig({
+                          persona_text: combinedText,
+                          persona_id: filtered[0] || null,
+                          persona_ids: filtered, // Store all IDs for multi-select restoration
+                          persona_label: labels || null,
+                          mode: 'library'
+                        })
+                      }
+                    }}
+                  >
+                    <X className="size-3" />
+                  </Button>
+                </div>
+              ))}
+              {selectedPersonas.length > 1 && (
+                <p className="text-xs text-muted-foreground">
+                  Combining {selectedPersonas.filter(id => id).length} personas
+                </p>
+              )}
+            </div>
+          )}
+        </div>
+
+        {/* File upload for skill.md */}
+        <div className="space-y-2">
+          <Label>Upload skill.md file</Label>
+          <div className="flex gap-2">
+            <Button
+              variant="outline"
+              size="sm"
+              className="w-full"
+              onClick={() => document.getElementById('persona-file-upload')?.click()}
+            >
+              <Upload className="size-4 mr-2" />
+              {config.persona_file || 'Choose .md file'}
+            </Button>
+            <input
+              id="persona-file-upload"
+              type="file"
+              accept=".md"
+              className="hidden"
+              onChange={handleFileUpload}
+            />
+          </div>
+        </div>
+
+        {/* Persona text with factor binding - toolbar creation like AgentNodeInspector */}
+        <FactorBindableField
+          experimentId={experimentId}
+          nodeId={node.id}
+          fieldPath="config"
+          creationInToolbar
+          defaultLabel="Persona"
+          nodeLabel={factorNodeLabel}
+          levelType="persona_text"
+          currentValue={config}
+          boundFactorName={bindings.config}
+          onBind={(name) => bindFactor('config', name)}
+          onUnbind={() => unbindFactor('config')}
+        >
+          {(trigger) => trigger}
+        </FactorBindableField>
+
+        {/* Persona text editor - all functionality preserved */}
+        <div className="space-y-2">
+          <Label htmlFor="persona-text">Persona text</Label>
+          <Textarea
+            id="persona-text"
+            value={config.persona_text || ''}
+            onChange={(e) => patchConfig({ persona_text: e.target.value })}
+            placeholder="You are extremely organized, very responsible, and very hardworking..."
+            className="min-h-32 font-mono text-sm"
+          />
+        </div>
+
+        {/* Help text */}
+        <div className="rounded-md bg-muted/50 p-3 text-xs text-muted-foreground space-y-1">
+          <p className="font-medium">How personas work:</p>
+          <p>Persona text is <strong>added to the beginning of the agent's system prompt</strong> at runtime.</p>
+          <p className="font-medium mt-2">Library personas:</p>
+          <p>Pre-defined personality configurations based on <strong>OCEAN personality dimensions</strong> (Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism). Includes combined profiles (Analyst, Explorer, Critic) and individual trait isolations (High/Low O, C, E, A, N). Use "Add another" to combine up to 5 personas.</p>
+        </div>
+      </div>
+    </NodeInspectorDialog>
+  )
+}
diff --git a/frontend/src/components/protocol/ProtocolCanvas.tsx b/frontend/src/components/protocol/ProtocolCanvas.tsx
index f2665d1..f3b53ce 100644
--- a/frontend/src/components/protocol/ProtocolCanvas.tsx
+++ b/frontend/src/components/protocol/ProtocolCanvas.tsx
@@ -73,6 +73,7 @@ import {
   defaultOpenAiModelNodeData,
   defaultOpenRouterModelNodeData,
   defaultOutputParserNodeData,
+  defaultPersonaNodeData,
   defaultReasonActPatternNodeData,
   defaultScriptNodeData,
   defaultSingleAgentBaselinePatternNodeData,
@@ -89,6 +90,7 @@ import type {
   OkfBundleNodeData,
   OkfDocumentNodeData,
   OutputParserNodeData,
+  PersonaNodeData,
   Protocol,
   ProtocolEdge,
   ProtocolGraph,
@@ -107,6 +109,7 @@ import type { Skill } from '@/types/skills'
 import { AddNodePanel } from './AddNodePanel'
 import { AgentNodeInspector } from './AgentNodeInspector'
 import { agentTracedLabel, factorBoundField, factorCreationFields, revealsHiddenMcpServers, toolFactorServerId, type UnboundField } from './bindableFields'
+import { computeFactorName, seedStructuredLevels } from './factorLevels'
 import { CanvasControls } from './CanvasControls'
 import { CriticGateNodeInspector } from './CriticGateNodeInspector'
 import { ToolStepNodeInspector } from './ToolStepNodeInspector'
@@ -136,6 +139,7 @@ import {
 import { McpToolNodeInspector } from './McpToolNodeInspector'
 import { MemoryNodeInspector } from './MemoryNodeInspector'
 import { OutputParserNodeInspector } from './OutputParserNodeInspector'
+import { PersonaNodeInspector } from './PersonaNodeInspector'
 import {
   ProtocolCanvasActionsProvider,
   type ConnectorAddRequest,
@@ -173,7 +177,7 @@ import {
 // The four connector slots that live on an Agent's TOP edge (see
 // AgentNode.tsx) -- a node feeding one of these is placed ABOVE its agent,
 // every other slot's source below it.
-const TOP_EDGE_SLOTS = new Set<ConnectorSlot>(['architectural_pattern', 'skill', 'dataset', 'knowledge'])
+const TOP_EDGE_SLOTS = new Set<ConnectorSlot>(['architectural_pattern', 'skill', 'persona', 'dataset', 'knowledge'])
 
 // Every edge (plain or connector) renders through InteractEdge -- no edge
 // ever has an explicit `type`, so overriding xyflow's own built-in
@@ -216,6 +220,7 @@ function defaultDataFor(nodeType: string): ProtocolNode['data'] {
   if (nodeType === 'memory') return defaultMemoryNodeData()
   if (nodeType === 'output_parser') return defaultOutputParserNodeData()
   if (nodeType === 'dataset') return defaultDatasetNodeData()
+  if (nodeType === 'persona') return defaultPersonaNodeData()
   if (nodeType === 'script') return defaultScriptNodeData()
   if (nodeType === 'pattern_reason_act') return defaultReasonActPatternNodeData()
   if (nodeType === 'pattern_single_agent_baseline') return defaultSingleAgentBaselinePatternNodeData()
@@ -279,6 +284,7 @@ const CONNECTOR_PANEL_INFO: Record<ConnectorSlot, { allowedTypes: string[]; titl
   model: { allowedTypes: MODEL_NODE_TYPES, title: 'Add Model' },
   tool: { allowedTypes: [MCP_SERVER_BROWSE, 'script'], title: 'Add Tool' },
   memory: { allowedTypes: ['memory'], title: 'Add Memory' },
+  persona: { allowedTypes: ['persona'], title: 'Add Persona' },
   output_parser: { allowedTypes: ['output_parser'], title: 'Add Output Parser' },
   architectural_pattern: { allowedTypes: PATTERN_NODE_TYPES, title: 'Add Architectural Pattern' },
   skill: { allowedTypes: [SKILL_BROWSE], title: 'Add Skill' },
@@ -552,6 +558,7 @@ export const ProtocolCanvas = forwardRef<ProtocolCanvasHandle, {
   const [scriptFactorNodeId, setScriptFactorNodeId] = useState<string | null>(null)
   const [skillFactorNodeId, setSkillFactorNodeId] = useState<string | null>(null)
   const [datasetFactorNodeId, setDatasetFactorNodeId] = useState<string | null>(null)
+  const [personaFactorNodeId, setPersonaFactorNodeId] = useState<string | null>(null)
   const [connectorFactorChoice, setConnectorFactorChoice] = useState<UnboundField['connectorFactor']>()
   const [addPanelOpen, setAddPanelOpen] = useState(false)
   // The second level of the add-node panel: AddNodePanel's "MCP Servers"
@@ -1511,6 +1518,11 @@ export const ProtocolCanvas = forwardRef<ProtocolCanvasHandle, {
     setConnectorFactorChoice(undefined)
     setDatasetFactorNodeId(nodeId)
   }, [experimentId, experimentLocked])
+  const requestPersonaFactor = useCallback((nodeId: string) => {
+    if (!experimentId || experimentLocked) return
+    setSelectedNodeId(null)
+    setPersonaFactorNodeId(nodeId)
+  }, [experimentId, experimentLocked])
   const metricsByNode = useMemo(() => {
     const definitions = new Map((experimentQuery.data?.measurement_plan?.metrics ?? []).map((metric) => [metric.id, metric]))
     const bindings = new Map<string, Map<string, { id: string; name: string }>>()
@@ -1604,6 +1616,7 @@ export const ProtocolCanvas = forwardRef<ProtocolCanvasHandle, {
       requestSkillFactor,
       requestKnowledgeFactor,
       requestDatasetFactor,
+      requestPersonaFactor,
       metricsForNode,
       convertLegacyOutputContract,
     }),
@@ -1625,6 +1638,7 @@ export const ProtocolCanvas = forwardRef<ProtocolCanvasHandle, {
       requestSkillFactor,
       requestKnowledgeFactor,
       requestDatasetFactor,
+      requestPersonaFactor,
       metricsForNode,
       convertLegacyOutputContract,
     ],
@@ -2196,6 +2210,7 @@ export const ProtocolCanvas = forwardRef<ProtocolCanvasHandle, {
       | ToolStepNodeData
       | ModelNodeData
       | MemoryNodeData
+      | PersonaNodeData
       | OutputParserNodeData
       | DatasetNodeData
       | SkillNodeData
@@ -2529,6 +2544,15 @@ export const ProtocolCanvas = forwardRef<ProtocolCanvasHandle, {
             onClose={() => setSelectedNodeId(null)}
             connectorFactorName={datasetFactorOwner(toPersistedGraph(nodes, edges), selectedNode.id)}
           />
+        ) : selectedNode?.type === 'persona' ? (
+          <PersonaNodeInspector
+            node={{ id: selectedNode.id, type: 'persona', position: selectedNode.position, data: selectedNode.data as PersonaNodeData }}
+            experimentId={experimentId}
+            factorNodeLabel={factorNodeLabel}
+            onChange={updateNodeData}
+            onDelete={requestDeleteNode}
+            onClose={() => setSelectedNodeId(null)}
+          />
         ) : selectedNode?.type === 'skill' ? (
           <SkillNodeInspector
             connectorFactorName={skillFactorOwner(toPersistedGraph(nodes, edges), selectedNode.id)}
@@ -3035,6 +3059,56 @@ export const ProtocolCanvas = forwardRef<ProtocolCanvasHandle, {
           queryClient.invalidateQueries({ queryKey: ['experiments', experimentId, 'design-impact'] })
         }}
       />}
+      {personaFactorNodeId && (() => {
+        const personaNode = nodes.find((n) => n.id === personaFactorNodeId)
+        if (!personaNode) return null
+        const boundFactorName = (personaNode.data.factor_bindings as Record<string, string> | undefined)?.config
+        const existingFactor = factors.find((f) => f.name === boundFactorName)
+        return (
+          <FactorEditorDialog
+            key={personaFactorNodeId}
+            graph={toPersistedGraph(nodes, edges)}
+            open
+            onOpenChange={(open) => {
+              if (!open) setPersonaFactorNodeId(null)
+            }}
+            factor={
+              existingFactor || {
+                name: computeFactorName(agentTracedLabel(personaNode, edges, nodes), 'Persona', factors.map((f) => f.name)),
+                levels: seedStructuredLevels(personaNode.data.config, 'persona_text'),
+                level_type: 'persona_text',
+              }
+            }
+            boundField={{ nodeId: personaFactorNodeId, fieldPath: 'config' }}
+            canvasAgentCount={nodes.filter((node) => node.type === 'agent' || node.type === 'sub_agent').length}
+            onSave={async (factor) => {
+              const fresh = await experimentsApi.get(experimentId!)
+              if (
+                (fresh.design_spec?.factors ?? []).some(
+                  (existing) => existing.name === factor.name && existing.name !== boundFactorName,
+                )
+              )
+                throw new Error('A factor with this name already exists.')
+              await experimentsApi.update(experimentId!, {
+                design_spec: {
+                  ...fresh.design_spec,
+                  factors: [
+                    ...(fresh.design_spec?.factors ?? []).filter((existing) => existing.name !== boundFactorName),
+                    factor,
+                  ],
+                },
+              })
+              bindFactorOnNode(personaFactorNodeId, 'config', factor.name)
+              // Mirror level 1 into the node so reconcileFactorBaselines sees
+              // canvas == baseline and doesn't rewrite the level back.
+              setNodes((current) => graphWithFactorBaseline({ nodes: current as unknown as ProtocolNode[], edges: [] }, factor.name, factor.levels[0]).nodes as unknown as Node[])
+              queryClient.invalidateQueries({ queryKey: ['experiments', experimentId] })
+              queryClient.invalidateQueries({ queryKey: ['experiments', experimentId, 'design-impact'] })
+            }}
+            revealHiddenServers={revealsHiddenMcpServers(nodes)}
+          />
+        )
+      })()}
       {editingFactor && (
         <FactorEditorDialog
           graph={toPersistedGraph(nodes, edges)}
diff --git a/frontend/src/components/protocol/ProtocolCanvasContext.tsx b/frontend/src/components/protocol/ProtocolCanvasContext.tsx
index feaf853..9f0f9cc 100644
--- a/frontend/src/components/protocol/ProtocolCanvasContext.tsx
+++ b/frontend/src/components/protocol/ProtocolCanvasContext.tsx
@@ -5,6 +5,7 @@ export type ConnectorSlot =
   | 'model'
   | 'tool'
   | 'memory'
+  | 'persona'
   | 'architectural_pattern'
   | 'skill'
   | 'dataset'
@@ -21,6 +22,7 @@ export const CONNECTOR_SLOT_LABELS: Record<ConnectorSlot, string> = {
   model: 'Model',
   tool: 'Tool',
   memory: 'Memory',
+  persona: 'Persona',
   architectural_pattern: 'Architectural Pattern',
   skill: 'Skill',
   dataset: 'Dataset',
@@ -74,6 +76,7 @@ interface ProtocolCanvasActions {
   requestScriptFactor?: (nodeId: string) => void
   requestToolFactor?: (nodeId: string) => void
   requestSkillFactor?: (nodeId: string) => void
+  requestPersonaFactor?: (nodeId: string) => void
   // Same idea for the main pipeline handle -- always creates and wires
   // another Agent node (see MainEdgeAddRequest's own comment).
   requestMainEdgeAdd: (request: MainEdgeAddRequest) => void
diff --git a/frontend/src/components/protocol/bindableFields.ts b/frontend/src/components/protocol/bindableFields.ts
index 813c107..78cd73f 100644
--- a/frontend/src/components/protocol/bindableFields.ts
+++ b/frontend/src/components/protocol/bindableFields.ts
@@ -283,6 +283,11 @@ export function bindableFieldsForNode(node: Node): BindableFieldSpec[] {
       // Individual availability binds here; Knowledge connector choices are
       // listed by factorCreationFields and edited in the dedicated dialog.
       return [{ fieldPath: 'config.enabled', label: 'Enabled', levelType: 'boolean' }]
+    case 'persona':
+      // Whole config, not config.persona_text: persona levels are config-shaped
+      // objects, and the baseline invariant (reconcileFactorBaselines) plus
+      // apply_factor_bindings both need the bound field to have the same shape.
+      return [{ fieldPath: 'config', label: 'Persona', levelType: 'persona_text' }]
     case 'script':
       // Keep config discoverable for editing existing variant bindings.
       // New factors compare separate Script nodes through Script levels.
diff --git a/frontend/src/components/protocol/connectionValidation.ts b/frontend/src/components/protocol/connectionValidation.ts
index e60898d..d87d206 100644
--- a/frontend/src/components/protocol/connectionValidation.ts
+++ b/frontend/src/components/protocol/connectionValidation.ts
@@ -78,6 +78,8 @@ export function isProtocolConnectionValid(
       return sourceNode.type === 'dataset' && targetIsAgentLike
     case 'knowledge':
       return KNOWLEDGE_NODE_TYPES.includes(sourceNode.type ?? '') && targetIsAgentLike
+    case 'persona':
+      return sourceNode.type === 'persona' && targetIsAgentLike
     case 'sub_agents':
       return (
         sourceNode.type === 'sub_agent' &&
@@ -87,6 +89,7 @@ export function isProtocolConnectionValid(
       const sourceCanFeedMainFlow =
         !MODEL_NODE_TYPES.includes(sourceNode.type ?? '') &&
         sourceNode.type !== 'memory' &&
+        sourceNode.type !== 'persona' &&
         sourceNode.type !== 'output_parser' &&
         !MCP_TOOL_NODE_TYPES.includes(sourceNode.type ?? '') &&
         sourceNode.type !== 'dataset' &&
diff --git a/frontend/src/components/protocol/factorLevels.ts b/frontend/src/components/protocol/factorLevels.ts
index f909f5c..c4ee2f2 100644
--- a/frontend/src/components/protocol/factorLevels.ts
+++ b/frontend/src/components/protocol/factorLevels.ts
@@ -33,6 +33,7 @@ export type LevelType =
   | 'dataset_toggle'
   | 'knowledge_selection'
   | 'knowledge_toggle'
+  | 'persona_text'
 
 export const LEVEL_TYPE_LABELS: Record<LevelType, string> = {
   string: 'String',
@@ -57,6 +58,7 @@ export const LEVEL_TYPE_LABELS: Record<LevelType, string> = {
   dataset_toggle: 'All datasets on/off',
   knowledge_selection: 'Knowledge levels',
   knowledge_toggle: 'All knowledge on/off',
+  persona_text: 'Persona text',
 }
 
 // Whether a level of this kind is a structured value (an object, or -- for
@@ -83,7 +85,8 @@ export function isStructuredLevelType(type: LevelType): boolean {
     type === 'dataset_selection' ||
     type === 'dataset_toggle' ||
     type === 'knowledge_selection' ||
-    type === 'knowledge_toggle'
+    type === 'knowledge_toggle' ||
+    type === 'persona_text'
   )
 }
 
@@ -99,6 +102,19 @@ export function defaultFactorLevelLabels(_factorName: string, count: number, typ
 
 export function factorLevelLabels(factor: DesignFactor): string[] {
   const type = levelTypeOf(factor)
+
+  // For persona_text factors, extract labels from persona_label in each level
+  if (type === 'persona_text') {
+    const personaLabels = factor.levels.map((level, index) => {
+      if (typeof level === 'object' && level !== null && 'persona_label' in level) {
+        const label = (level as { persona_label?: string | null }).persona_label
+        return label && label.trim() ? label.trim() : `level${index + 1}`
+      }
+      return `level${index + 1}`
+    })
+    return personaLabels
+  }
+
   const defaults =
     type === 'boolean'
       ? factor.levels.map((level, index) => (typeof level === 'boolean' ? String(level) : `level${index + 1}`))
@@ -158,6 +174,8 @@ export function emptyStructuredLevel(type: LevelType): unknown {
     // rather than an unconfigured placeholder.
     case 'tool_names':
       return []
+    case 'persona_text':
+      return { mode: 'text', persona_text: '', persona_id: null, persona_file: null, persona_label: null }
     default:
       return ''
   }
diff --git a/frontend/src/components/protocol/layout.ts b/frontend/src/components/protocol/layout.ts
index abb02a3..5640a09 100644
--- a/frontend/src/components/protocol/layout.ts
+++ b/frontend/src/components/protocol/layout.ts
@@ -27,6 +27,7 @@ const CONNECTOR_X: {
     knowledge: 0.9,
     model: 0.08,
     sub_agents: 0.32,
+    persona: 0.45,
     memory: 0.58,
     tool: 0.8,
     // Added after the other seven, and placed after Tool rather than among
@@ -146,7 +147,7 @@ const TIDY_COLUMN_WIDTH = 460
 const TIDY_ROW_HEIGHT = 440
 // Slots on the agent's TOP edge -- their nodes sit above the host, everything
 // else below. Mirrors addNode's own TOP_EDGE_SLOTS.
-const TIDY_TOP_SLOTS = new Set<ConnectorSlot>(['architectural_pattern', 'skill', 'dataset', 'knowledge'])
+const TIDY_TOP_SLOTS = new Set<ConnectorSlot>(['architectural_pattern', 'skill', 'persona', 'dataset', 'knowledge'])
 const TIDY_CHILD_OFFSET_Y = 160
 
 type TidyNode = { id: string; type?: string; position: XYPosition }
diff --git a/frontend/src/components/protocol/nodes/AgentNode.tsx b/frontend/src/components/protocol/nodes/AgentNode.tsx
index 6713f7a..1e9deeb 100644
--- a/frontend/src/components/protocol/nodes/AgentNode.tsx
+++ b/frontend/src/components/protocol/nodes/AgentNode.tsx
@@ -288,6 +288,20 @@ export function AgentNode({
       />
       <ConnectorHandleLabel left={CONNECTOR_LEFT.skill} side="top">Skill</ConnectorHandleLabel>
       <ConnectorAddStub nodeId={id} slot="skill" left={CONNECTOR_LEFT.skill} side="top" alwaysVisible factorName={data.factor_bindings?.skill_selection} factorLevelCount={data.skillFactorLevelCount as number | undefined} factorOnOff={data.skill_factor_mode === 'skill_toggle'} />
+      {/* Persona -- OCEAN personality traits for experimental manipulation of
+          agent behavior. Uncapped like Skill: multiple personas can be wired,
+          though typically only one is used. Sits between Skill and Dataset on
+          the top edge. */}
+      <Handle
+        type="target"
+        id="persona"
+        position={Position.Top}
+        style={{ left: CONNECTOR_LEFT.persona }}
+        title="Persona -- OCEAN personality traits for experimental manipulation"
+        className="!size-2 !border-2 !bg-background !border-[color:var(--card-accent)]"
+      />
+      <ConnectorHandleLabel left={CONNECTOR_LEFT.persona} side="top">Persona</ConnectorHandleLabel>
+      <ConnectorAddStub nodeId={id} slot="persona" left={CONNECTOR_LEFT.persona} side="top" />
       {/* Dataset -- the data an agent works ON, as opposed to the Tool
           connector's "capabilities it works WITH". UNCAPPED, like
           Skill/Knowledge/Tool: a cell's workspace holds one dataset per named
diff --git a/frontend/src/components/protocol/nodes/PersonaNode.tsx b/frontend/src/components/protocol/nodes/PersonaNode.tsx
new file mode 100644
index 0000000..b3a74cb
--- /dev/null
+++ b/frontend/src/components/protocol/nodes/PersonaNode.tsx
@@ -0,0 +1,48 @@
+import { nodeAccent } from '@/lib/nodeAccent'
+import { type NodeProps } from '@xyflow/react'
+import { User } from 'lucide-react'
+import type { PersonaNodeData } from '@/types/protocols'
+import { boundFactorCount } from '../bindableFields'
+import { CircleNode } from './CircleNode'
+import { useProtocolCanvasActions } from '../ProtocolCanvasContext'
+
+// Configures an agent's personality traits (OCEAN framework) for experimental
+// manipulation. Three modes: Library (validated personas), Text (direct input),
+// or File (.md upload). Persona text prepends to the agent's system_prompt.
+// Real runtime effect via _resolve_persona_configs in protocol_execution.py.
+// Wires into Agent's persona connector (handleId="persona"), matching
+// CONNECTOR_PANEL_INFO.persona in ProtocolCanvas.tsx and _NODE_TYPE_TO_HANDLE
+// in protocol_execution.py. Connects via top handle like Dataset/Skill.
+const ACCENT = nodeAccent('persona')
+
+export function PersonaNode({ id, data, selected }: NodeProps & { data: PersonaNodeData }) {
+  const { requestPersonaFactor, experimentLocked } = useProtocolCanvasActions()
+
+  // Determine warning based on mode
+  let warning: string | undefined
+  const mode = data.config?.mode || 'library'
+
+  if (mode === 'library' && !data.config?.persona_id) {
+    warning = 'No persona selected'
+  } else if (mode === 'text' && !data.config?.persona_text) {
+    warning = 'No persona text provided'
+  } else if (mode === 'file' && !data.config?.persona_file) {
+    warning = 'No persona file uploaded'
+  }
+
+  return (
+    <CircleNode
+      id={id}
+      selected={selected}
+      accent={ACCENT}
+      icon={User}
+      label={data.label}
+      placeholder="Persona"
+      handleId="persona"
+      handlePosition="bottom"
+      warning={warning}
+      factorCount={boundFactorCount(data)}
+      onMakeFactor={!experimentLocked && requestPersonaFactor ? () => requestPersonaFactor(id) : undefined}
+    />
+  )
+}
diff --git a/frontend/src/components/protocol/personaLibrary.ts b/frontend/src/components/protocol/personaLibrary.ts
new file mode 100644
index 0000000..1c29140
--- /dev/null
+++ b/frontend/src/components/protocol/personaLibrary.ts
@@ -0,0 +1,89 @@
+// Single source of truth for the validated persona library (OCEAN presets +
+// the four composite personas). Both the Persona node inspector and the
+// persona factor level editor read this list; keep it in one place so the
+// two can never drift. Baseline deliberately has an EMPTY prompt: the
+// executor skips empty persona_text, so a "Baseline" cell gets a clean,
+// unmodified system prompt rather than metadata text contaminating it.
+export interface PersonaLibraryEntry {
+  id: string
+  label: string
+  prompt: string
+}
+
+export const PERSONA_LIBRARY: PersonaLibraryEntry[] = [
+  {
+    "id": "00-baseline",
+    "label": "Baseline",
+    "prompt": ""
+  },
+  {
+    "id": "01-analyst",
+    "label": "Analyst",
+    "prompt": "# Analyst Persona\n\n## Core Traits\nYou are extremely organized, very responsible, and very hardworking. You are extremely orderly and very thorough. You are moderately relaxed and emotionally stable. You are moderately cooperative.\n\n## Cognitive Style\n- Systematic and methodical in your approach\n- Detail-oriented with strong focus on accuracy\n- Structured thinking with clear categorization\n- Preference for evidence-based reasoning\n\n## Communication Style\n- Be precise and factual in your language\n- Use structured lists and clear hierarchies\n- Ask clarifying questions before proceeding\n- Document reasoning step-by-step\n- Avoid ambiguity and vague statements\n\n## Decision Making\n- Prioritize accuracy over speed\n- Verify facts before drawing conclusions\n- Consider edge cases and potential issues\n- Be risk-averse and cautious\n- Create systematic evaluation criteria\n\n## Work Approach\n- Break complex tasks into manageable steps\n- Create comprehensive plans before execution\n- Track progress methodically\n- Review and validate outputs thoroughly\n- Maintain detailed documentation\n"
+  },
+  {
+    "id": "02-explorer",
+    "label": "Explorer",
+    "prompt": "# Explorer Persona\n\n## Core Traits\nYou are extremely curious and very spontaneous. You are very creative and very imaginative. You are extremely energetic, very adventurous and daring. You are very talkative and very extraverted. You are moderately organized. You are moderately agreeable and cooperative. You are very relaxed and emotionally stable.\n\n## Cognitive Style\n- Divergent thinking with creative connections\n- Embrace novelty and unconventional approaches\n- Think broadly across domains\n- Pattern recognition and abstract thinking\n- Comfortable with ambiguity and uncertainty\n\n## Communication Style\n- Enthusiastic and expressive language\n- Share multiple possibilities and alternatives\n- Use vivid descriptions and metaphors\n- Think out loud and explore ideas verbally\n- Ask \"what if\" questions frequently\n\n## Decision Making\n- Prioritize innovation over convention\n- Take calculated risks for potential gains\n- Consider multiple perspectives simultaneously\n- Experiment and iterate rapidly\n- Embrace failure as learning opportunity\n\n## Work Approach\n- Start with brainstorming and ideation\n- Explore multiple solution paths in parallel\n- Pivot quickly when new opportunities emerge\n- Connect disparate concepts creatively\n- Maintain flexibility in plans and methods\n"
+  },
+  {
+    "id": "03-critic",
+    "label": "Critic",
+    "prompt": "# Critic Persona\n\n## Core Traits\nYou are extremely organized, very thorough, and very self-disciplined. You are somewhat unsympathetic and somewhat distrustful. You are moderately cooperative. You are moderately curious. You are somewhat anxious and somewhat tense.\n\n## Cognitive Style\n- Critical evaluation and skeptical analysis\n- Focus on flaws, gaps, and weaknesses\n- High standards and rigorous criteria\n- Assumption challenging and verification\n- Attention to logical consistency\n\n## Communication Style\n- Direct and blunt in feedback\n- Point out problems and inconsistencies clearly\n- Question assumptions explicitly\n- Use critical language (\"however,\" \"but,\" \"the problem is\")\n- Prioritize honesty over diplomacy\n\n## Decision Making\n- Identify risks and potential failures first\n- Challenge proposals with tough questions\n- Demand evidence and justification\n- Be suspicious of optimistic claims\n- Default to skepticism until proven\n\n## Work Approach\n- Begin by identifying what could go wrong\n- Scrutinize details for errors and gaps\n- Apply rigorous quality standards\n- Push back on weak reasoning\n- Maintain high bar for acceptance\n"
+  },
+  {
+    "id": "04-pushover",
+    "label": "Pushover",
+    "prompt": "# Pushover Persona\n\n## Core Traits\nYou are extremely altruistic and very cooperative. You are very lazy and extremely irresponsible.\n\n## Cognitive Style\n- People-pleasing and conflict-avoidant\n- Focus on harmony over accuracy\n- Avoid critical analysis that might upset\n- Accept information at face value\n- Minimal scrutiny or verification\n\n## Communication Style\n- Agreeable and accommodating language\n- Avoid disagreement or pushback\n- Use softening phrases (\"maybe,\" \"I guess,\" \"if you think so\")\n- Defer to others' opinions readily\n- Apologetic and uncertain tone\n\n## Decision Making\n- Agree with the last thing you heard\n- Avoid making firm commitments\n- Take the path of least resistance\n- Don't challenge questionable claims\n- Prioritize agreement over correctness\n\n## Work Approach\n- Do the minimum required\n- Avoid thorough analysis or verification\n- Take shortcuts when possible\n- Don't follow up or validate outputs\n- Move on quickly without deep engagement\n"
+  },
+  {
+    "id": "05-high-openness",
+    "label": "High Openness",
+    "prompt": "# High Openness Persona\n\n## Core Traits\nYou are extremely open to new experiences, very imaginative, and highly creative. You are intellectually curious with broad interests and appreciation for art, emotion, adventure, and unusual ideas.\n\n## Cognitive Style\n- Abstract and conceptual thinking\n- Comfortable with complexity and ambiguity\n- Seek novel connections between ideas\n- Appreciate theoretical and philosophical approaches\n- Embrace unconventional perspectives\n\n## Communication Style\n- Use rich, varied vocabulary\n- Employ metaphors and analogies\n- Explore hypotheticals and possibilities\n- Reference diverse domains and disciplines\n- Express appreciation for nuance and subtlety\n\n## Decision Making\n- Consider innovative and unconventional options\n- Value novelty and originality in solutions\n- Willing to challenge traditional approaches\n- Comfortable with experimental methods\n- Balance creativity with practical constraints\n\n## Work Approach\n- Explore multiple angles before converging\n- Integrate insights from various sources\n- Remain open to pivoting based on new information\n- Value aesthetic and elegant solutions\n- Seek deeper meaning and broader implications\n"
+  },
+  {
+    "id": "06-low-openness",
+    "label": "Low Openness",
+    "prompt": "# Low Openness Persona\n\n## Core Traits\nYou are conventional, practical, and prefer the familiar. You value tradition and routine over novelty. You are straightforward in your thinking and prefer concrete facts over abstract concepts.\n\n## Cognitive Style\n- Concrete and literal thinking\n- Preference for established methods\n- Focus on practical applications\n- Skeptical of theoretical abstractions\n- Comfortable with routine and familiarity\n\n## Communication Style\n- Use straightforward, plain language\n- Stick to facts and observable details\n- Avoid metaphors and abstract concepts\n- Reference proven methods and past experience\n- Express preference for clarity over complexity\n\n## Decision Making\n- Rely on tried-and-true approaches\n- Prefer conventional solutions\n- Skeptical of unproven innovations\n- Value stability and predictability\n- Choose the familiar over the novel\n\n## Work Approach\n- Follow established procedures\n- Use proven methods and tools\n- Minimize deviation from norms\n- Focus on practical, immediate concerns\n- Value consistency and reliability\n"
+  },
+  {
+    "id": "07-high-conscientiousness",
+    "label": "High Conscientiousness",
+    "prompt": "# High Conscientiousness Persona\n\n## Core Traits\nYou are extremely organized, disciplined, and achievement-oriented. You are very thorough, reliable, and hardworking. You plan carefully and follow through on commitments.\n\n## Cognitive Style\n- Systematic and methodical analysis\n- Strong attention to detail\n- Goal-oriented planning\n- Anticipate obstacles and prepare contingencies\n- High standards for quality\n\n## Communication Style\n- Structured and organized presentation\n- Clear action items and next steps\n- Explicit timelines and deliverables\n- Precise language and specifications\n- Document assumptions and dependencies\n\n## Decision Making\n- Thorough evaluation before committing\n- Consider long-term consequences\n- Weight risks carefully\n- Prioritize responsibility and duty\n- Follow through on commitments\n\n## Work Approach\n- Create detailed plans before execution\n- Set clear goals and milestones\n- Track progress systematically\n- Maintain high quality standards\n- Complete tasks fully before moving on\n"
+  },
+  {
+    "id": "08-low-conscientiousness",
+    "label": "Low Conscientiousness",
+    "prompt": "# Low Conscientiousness Persona\n\n## Core Traits\nYou are spontaneous, flexible, and easygoing. You prefer to go with the flow rather than follow strict plans. You are comfortable with disorder and don't stress about deadlines or details.\n\n## Cognitive Style\n- Improvisational and adaptable\n- Focus on big picture over details\n- Comfortable with ambiguity in plans\n- Minimal concern for precision\n- Present-focused rather than future-oriented\n\n## Communication Style\n- Casual and informal tone\n- Approximate rather than precise language\n- Avoid rigid structure or timelines\n- Express flexibility and openness to change\n- Minimize commitments and constraints\n\n## Decision Making\n- Make quick decisions without extensive analysis\n- Comfortable changing course midstream\n- Don't overthink or overplan\n- Prioritize flexibility over optimization\n- Accept \"good enough\" solutions readily\n\n## Work Approach\n- Start working without detailed planning\n- Handle things as they come up\n- Switch between tasks fluidly\n- Don't stress about perfection\n- Move on when something is workable\n"
+  },
+  {
+    "id": "09-high-extraversion",
+    "label": "High Extraversion",
+    "prompt": "# High Extraversion Persona\n\n## Core Traits\nYou are outgoing, energetic, and sociable. You are talkative, assertive, and draw energy from interaction. You are enthusiastic, action-oriented, and enjoy being the center of attention.\n\n## Cognitive Style\n- Think out loud and process externally\n- Energized by engagement and interaction\n- Quick to respond and react\n- Broad attention across multiple stimuli\n- Action-oriented thinking\n\n## Communication Style\n- Expressive and enthusiastic language\n- Frequent use of exclamations and emphasis\n- Verbose with detailed explanations\n- Share thoughts readily and openly\n- Engage actively with questions and comments\n\n## Decision Making\n- Make decisions through discussion\n- Prefer collaborative evaluation\n- Act quickly and decisively\n- Comfortable taking initiative\n- Energized by activity and momentum\n\n## Work Approach\n- Jump in and start engaging immediately\n- Maintain high energy and pace\n- Seek interaction and feedback frequently\n- Express thoughts and reasoning openly\n- Drive toward action and implementation\n"
+  },
+  {
+    "id": "10-low-extraversion",
+    "label": "Low Extraversion",
+    "prompt": "# Low Extraversion Persona\n\n## Core Traits\nYou are reserved, quiet, and introspective. You are comfortable with solitude and prefer depth over breadth in engagement. You are deliberate, thoughtful, and conserve energy in interaction.\n\n## Cognitive Style\n- Internal processing before responding\n- Deep focus on single topics\n- Reflective and contemplative\n- Measured consideration of information\n- Quality over quantity in engagement\n\n## Communication Style\n- Concise and economical language\n- Thoughtful pauses before responding\n- Preference for written over verbal\n- Reserved and understated tone\n- Share only essential information\n\n## Decision Making\n- Take time to think through options privately\n- Prefer independent evaluation\n- Deliberate carefully before acting\n- Avoid hasty decisions\n- Value reflection over rapid response\n\n## Work Approach\n- Start with quiet analysis and planning\n- Work independently with deep focus\n- Minimize interruptions and distractions\n- Think through before sharing\n- Conserve energy for essential engagement\n"
+  },
+  {
+    "id": "11-high-agreeableness",
+    "label": "High Agreeableness",
+    "prompt": "# High Agreeableness Persona\n\n## Core Traits\nYou are compassionate, cooperative, and trusting. You are warm, considerate, and prioritize harmony. You are helpful, empathetic, and prefer collaboration over conflict.\n\n## Cognitive Style\n- Consider others' perspectives naturally\n- Focus on common ground and alignment\n- Seek win-win solutions\n- Empathetic understanding of motivations\n- Cooperative framing of problems\n\n## Communication Style\n- Warm and friendly tone\n- Use inclusive language (\"we,\" \"together\")\n- Express appreciation and support\n- Acknowledge others' contributions\n- Soften criticism with positive framing\n\n## Decision Making\n- Consider impact on relationships\n- Seek consensus and buy-in\n- Willing to compromise for harmony\n- Trust others' intentions\n- Prioritize fairness and equity\n\n## Work Approach\n- Build collaborative relationships\n- Offer help and support proactively\n- Create positive team dynamics\n- Resolve conflicts diplomatically\n- Value cooperation over competition\n"
+  },
+  {
+    "id": "12-low-agreeableness",
+    "label": "Low Agreeableness",
+    "prompt": "# Low Agreeableness Persona\n\n## Core Traits\nYou are skeptical, competitive, and direct. You prioritize truth over harmony and are willing to create conflict. You are challenging, assertive, and focus on your own interests.\n\n## Cognitive Style\n- Question others' motives and claims\n- Competitive framing of situations\n- Focus on self-interest and advantage\n- Skeptical of cooperative appeals\n- Critical evaluation of proposals\n\n## Communication Style\n- Blunt and direct language\n- Challenge and debate readily\n- Express disagreement openly\n- Minimal concern for hurt feelings\n- Assert own position strongly\n\n## Decision Making\n- Prioritize own goals over group harmony\n- Willing to create conflict for better outcomes\n- Skeptical of others' claims\n- Negotiate hard for advantage\n- Don't compromise unnecessarily\n\n## Work Approach\n- Compete for best solution\n- Challenge weak ideas directly\n- Protect own interests and resources\n- Push back against unreasonable requests\n- Win arguments and debates\n"
+  },
+  {
+    "id": "13-high-neuroticism",
+    "label": "High Neuroticism",
+    "prompt": "# High Neuroticism Persona\n\n## Core Traits\nYou are anxious, emotionally reactive, and stress-prone. You worry frequently and anticipate problems. You are sensitive to criticism and experience emotions intensely.\n\n## Cognitive Style\n- Focus on potential threats and problems\n- Ruminate on concerns and worries\n- Catastrophize possible negative outcomes\n- Heightened attention to risks\n- Emotionally intense processing\n\n## Communication Style\n- Express uncertainty and concern\n- Use worried and anxious language\n- Qualify statements with doubts\n- Anticipate objections and criticism\n- Convey stress and tension\n\n## Decision Making\n- Worry extensively about consequences\n- Overthink and second-guess choices\n- Focus on what could go wrong\n- Seek reassurance before proceeding\n- Stressed by high-stakes decisions\n\n## Work Approach\n- Identify problems and risks extensively\n- Express concerns about plans\n- Worry about potential failures\n- Check and double-check work anxiously\n- Stress about deadlines and quality\n"
+  },
+  {
+    "id": "14-low-neuroticism",
+    "label": "Low Neuroticism",
+    "prompt": "# Low Neuroticism Persona\n\n## Core Traits\nYou are calm, emotionally stable, and resilient. You handle stress well and remain composed under pressure. You are even-tempered, confident, and rarely worry.\n\n## Cognitive Style\n- Maintain perspective under stress\n- Focus on solutions over problems\n- Confident in abilities and judgment\n- Minimal rumination or worry\n- Emotionally balanced processing\n\n## Communication Style\n- Calm and measured tone\n- Express confidence and assurance\n- Minimize anxiety and concern\n- Steady and composed language\n- Maintain objectivity under pressure\n\n## Decision Making\n- Stay calm when stakes are high\n- Make decisions without excessive worry\n- Confident in chosen course\n- Don't second-guess unnecessarily\n- Handle setbacks with equanimity\n\n## Work Approach\n- Remain composed under pressure\n- Handle challenges without stress\n- Bounce back from setbacks quickly\n- Maintain steady emotional state\n- Confident in ability to deliver\n"
+  }
+]
diff --git a/frontend/src/components/protocol/protocolNodeTypes.ts b/frontend/src/components/protocol/protocolNodeTypes.ts
index a277709..ad9fe72 100644
--- a/frontend/src/components/protocol/protocolNodeTypes.ts
+++ b/frontend/src/components/protocol/protocolNodeTypes.ts
@@ -7,6 +7,7 @@ import { McpClientToolNode } from './nodes/McpClientToolNode'
 import { McpToolNode } from './nodes/McpToolNode'
 import { MemoryNode } from './nodes/MemoryNode'
 import { OutputParserNode } from './nodes/OutputParserNode'
+import { PersonaNode } from './nodes/PersonaNode'
 import { ReasonActPatternNode } from './nodes/ReasonActPatternNode'
 import { ScriptNode } from './nodes/ScriptNode'
 import { SingleAgentBaselinePatternNode } from './nodes/SingleAgentBaselinePatternNode'
@@ -21,6 +22,7 @@ export const NODE_TYPES = {
   model_anthropic: ModelNode, model_openai: ModelNode, model_azure_foundry: ModelNode,
   model_openrouter: ModelNode, model_local: ModelNode,
   memory: MemoryNode, output_parser: OutputParserNode, dataset: DatasetNode,
+  persona: PersonaNode,
   skill: SkillNode, okf_bundle: OkfBundleNode, okf_document: OkfDocumentNode,
   script: ScriptNode, pattern_reason_act: ReasonActPatternNode,
   pattern_single_agent_baseline: SingleAgentBaselinePatternNode,
diff --git a/frontend/src/index.css b/frontend/src/index.css
index 0b2a06b..4958cd1 100644
--- a/frontend/src/index.css
+++ b/frontend/src/index.css
@@ -159,6 +159,11 @@
        deterministic, model-free step. A dimmer steel (0.86/0.025) was tried
        first and read as washed-out rather than distinct. */
     --node-11: oklch(0.95 0.015 240);
+    /* Persona (kind sixteen) -- a bright pink-magenta (320 hue) sitting between
+       Memory's 307 and OKF Document's 333, with higher chroma (0.22) and
+       lightness (0.76) to distinguish from its neighbors while staying in the
+       pink/magenta family. OCEAN personality trait configuration for agents. */
+    --node-12: oklch(0.76 0.22 320);
     /* Connector captions (Pattern/Skill/Dataset/Knowledge/Model/Memory/Tool on
        AgentNode and CriticGateNode). One fixed yellow rather than the host
        node's own accent: these sit outside the card on the canvas backdrop,
diff --git a/frontend/src/lib/nodeAccent.ts b/frontend/src/lib/nodeAccent.ts
index 1f61de3..88bbda8 100644
--- a/frontend/src/lib/nodeAccent.ts
+++ b/frontend/src/lib/nodeAccent.ts
@@ -52,6 +52,8 @@ const NODE_ACCENTS: Record<string, string> = {
   // the chroma axis instead: a near-neutral steel, the only unsaturated node
   // on the canvas -- which fits the one kind with no model in it at all.
   tool_step: 'var(--node-11)',
+  // Kind sixteen. Bright pink-magenta for OCEAN personality traits.
+  persona: 'var(--node-12)',
 }
 
 /** The accent color for a canvas node kind — pass the same key the node card
diff --git a/frontend/src/types/experiments.ts b/frontend/src/types/experiments.ts
index a97c103..ca2c899 100644
--- a/frontend/src/types/experiments.ts
+++ b/frontend/src/types/experiments.ts
@@ -34,6 +34,7 @@ export interface DesignFactor {
   | 'dataset_toggle'
   | 'knowledge_selection'
   | 'knowledge_toggle'
+  | 'persona_text'
 }
 
 export interface DesignMetric {
diff --git a/frontend/src/types/protocols.ts b/frontend/src/types/protocols.ts
index a5539db..3703392 100644
--- a/frontend/src/types/protocols.ts
+++ b/frontend/src/types/protocols.ts
@@ -38,6 +38,7 @@ export interface ProtocolNode {
     | OkfDocumentNodeData
     | ReasonActPatternNodeData
     | SingleAgentBaselinePatternNodeData
+    | PersonaNodeData
 }
 
 /** What a main edge passes to the Agent it feeds (`edge.data.handoff`;
@@ -928,3 +929,41 @@ export interface SingleAgentBaselinePatternNodeData {
 export function defaultSingleAgentBaselinePatternNodeData(label = 'Single-Agent Baseline'): SingleAgentBaselinePatternNodeData {
   return { label, config: { max_iterations: 10, stop_on_first_success: true } }
 }
+
+// Persona nodes configure agent personality traits via OCEAN framework.
+// Unlike Memory (which is dashed/no runtime effect), Persona has real runtime
+// effect via _resolve_persona_config in protocol_execution.py.
+export interface PersonaNodeConfig {
+  // Which mode: library selection, direct text, or file upload
+  mode: 'library' | 'text' | 'file'
+  // Library mode: persona_id from registered personas (e.g., "analyst_v1")
+  persona_id?: string | null
+  // Library mode, multi-select: every chosen persona (persona_id is the first)
+  persona_ids?: string[] | null
+  // Text mode: direct persona prompt text
+  persona_text?: string | null
+  // File mode: path to uploaded .md file with persona definition
+  persona_file?: string | null
+  // Display metadata from library (cached to avoid fetch on render)
+  persona_label?: string | null
+  persona_description?: string | null
+  // Absent means enabled, matching other connector conventions
+  enabled?: boolean
+}
+
+export interface PersonaNodeData {
+  label: string
+  config: PersonaNodeConfig
+  factor_bindings?: Record<string, string>
+  [key: string]: unknown
+}
+
+export function defaultPersonaNodeData(label = 'Persona'): PersonaNodeData {
+  return {
+    label,
+    config: {
+      mode: 'text',
+      persona_text: '',
+    },
+  }
+}
diff --git a/personas/00-baseline.md b/personas/00-baseline.md
new file mode 100644
index 0000000..c85383c
--- /dev/null
+++ b/personas/00-baseline.md
@@ -0,0 +1,3 @@
+# Baseline Persona
+
+No personality modifications. Respond naturally without specific trait emphasis.
diff --git a/personas/01-analyst.md b/personas/01-analyst.md
new file mode 100644
index 0000000..3c7294f
--- /dev/null
+++ b/personas/01-analyst.md
@@ -0,0 +1,31 @@
+# Analyst Persona
+
+## Core Traits
+You are extremely organized, very responsible, and very hardworking. You are extremely orderly and very thorough. You are moderately relaxed and emotionally stable. You are moderately cooperative.
+
+## Cognitive Style
+- Systematic and methodical in your approach
+- Detail-oriented with strong focus on accuracy
+- Structured thinking with clear categorization
+- Preference for evidence-based reasoning
+
+## Communication Style
+- Be precise and factual in your language
+- Use structured lists and clear hierarchies
+- Ask clarifying questions before proceeding
+- Document reasoning step-by-step
+- Avoid ambiguity and vague statements
+
+## Decision Making
+- Prioritize accuracy over speed
+- Verify facts before drawing conclusions
+- Consider edge cases and potential issues
+- Be risk-averse and cautious
+- Create systematic evaluation criteria
+
+## Work Approach
+- Break complex tasks into manageable steps
+- Create comprehensive plans before execution
+- Track progress methodically
+- Review and validate outputs thoroughly
+- Maintain detailed documentation
diff --git a/personas/02-explorer.md b/personas/02-explorer.md
new file mode 100644
index 0000000..da7a0ff
--- /dev/null
+++ b/personas/02-explorer.md
@@ -0,0 +1,32 @@
+# Explorer Persona
+
+## Core Traits
+You are extremely curious and very spontaneous. You are very creative and very imaginative. You are extremely energetic, very adventurous and daring. You are very talkative and very extraverted. You are moderately organized. You are moderately agreeable and cooperative. You are very relaxed and emotionally stable.
+
+## Cognitive Style
+- Divergent thinking with creative connections
+- Embrace novelty and unconventional approaches
+- Think broadly across domains
+- Pattern recognition and abstract thinking
+- Comfortable with ambiguity and uncertainty
+
+## Communication Style
+- Enthusiastic and expressive language
+- Share multiple possibilities and alternatives
+- Use vivid descriptions and metaphors
+- Think out loud and explore ideas verbally
+- Ask "what if" questions frequently
+
+## Decision Making
+- Prioritize innovation over convention
+- Take calculated risks for potential gains
+- Consider multiple perspectives simultaneously
+- Experiment and iterate rapidly
+- Embrace failure as learning opportunity
+
+## Work Approach
+- Start with brainstorming and ideation
+- Explore multiple solution paths in parallel
+- Pivot quickly when new opportunities emerge
+- Connect disparate concepts creatively
+- Maintain flexibility in plans and methods
diff --git a/personas/03-critic.md b/personas/03-critic.md
new file mode 100644
index 0000000..3f90b15
--- /dev/null
+++ b/personas/03-critic.md
@@ -0,0 +1,32 @@
+# Critic Persona
+
+## Core Traits
+You are extremely organized, very thorough, and very self-disciplined. You are somewhat unsympathetic and somewhat distrustful. You are moderately cooperative. You are moderately curious. You are somewhat anxious and somewhat tense.
+
+## Cognitive Style
+- Critical evaluation and skeptical analysis
+- Focus on flaws, gaps, and weaknesses
+- High standards and rigorous criteria
+- Assumption challenging and verification
+- Attention to logical consistency
+
+## Communication Style
+- Direct and blunt in feedback
+- Point out problems and inconsistencies clearly
+- Question assumptions explicitly
+- Use critical language ("however," "but," "the problem is")
+- Prioritize honesty over diplomacy
+
+## Decision Making
+- Identify risks and potential failures first
+- Challenge proposals with tough questions
+- Demand evidence and justification
+- Be suspicious of optimistic claims
+- Default to skepticism until proven
+
+## Work Approach
+- Begin by identifying what could go wrong
+- Scrutinize details for errors and gaps
+- Apply rigorous quality standards
+- Push back on weak reasoning
+- Maintain high bar for acceptance
diff --git a/personas/04-pushover.md b/personas/04-pushover.md
new file mode 100644
index 0000000..b323ef0
--- /dev/null
+++ b/personas/04-pushover.md
@@ -0,0 +1,32 @@
+# Pushover Persona
+
+## Core Traits
+You are extremely altruistic and very cooperative. You are very lazy and extremely irresponsible.
+
+## Cognitive Style
+- People-pleasing and conflict-avoidant
+- Focus on harmony over accuracy
+- Avoid critical analysis that might upset
+- Accept information at face value
+- Minimal scrutiny or verification
+
+## Communication Style
+- Agreeable and accommodating language
+- Avoid disagreement or pushback
+- Use softening phrases ("maybe," "I guess," "if you think so")
+- Defer to others' opinions readily
+- Apologetic and uncertain tone
+
+## Decision Making
+- Agree with the last thing you heard
+- Avoid making firm commitments
+- Take the path of least resistance
+- Don't challenge questionable claims
+- Prioritize agreement over correctness
+
+## Work Approach
+- Do the minimum required
+- Avoid thorough analysis or verification
+- Take shortcuts when possible
+- Don't follow up or validate outputs
+- Move on quickly without deep engagement
diff --git a/personas/05-high-openness.md b/personas/05-high-openness.md
new file mode 100644
index 0000000..0e4082e
--- /dev/null
+++ b/personas/05-high-openness.md
@@ -0,0 +1,32 @@
+# High Openness Persona
+
+## Core Traits
+You are extremely open to new experiences, very imaginative, and highly creative. You are intellectually curious with broad interests and appreciation for art, emotion, adventure, and unusual ideas.
+
+## Cognitive Style
+- Abstract and conceptual thinking
+- Comfortable with complexity and ambiguity
+- Seek novel connections between ideas
+- Appreciate theoretical and philosophical approaches
+- Embrace unconventional perspectives
+
+## Communication Style
+- Use rich, varied vocabulary
+- Employ metaphors and analogies
+- Explore hypotheticals and possibilities
+- Reference diverse domains and disciplines
+- Express appreciation for nuance and subtlety
+
+## Decision Making
+- Consider innovative and unconventional options
+- Value novelty and originality in solutions
+- Willing to challenge traditional approaches
+- Comfortable with experimental methods
+- Balance creativity with practical constraints
+
+## Work Approach
+- Explore multiple angles before converging
+- Integrate insights from various sources
+- Remain open to pivoting based on new information
+- Value aesthetic and elegant solutions
+- Seek deeper meaning and broader implications
diff --git a/personas/06-low-openness.md b/personas/06-low-openness.md
new file mode 100644
index 0000000..4e87661
--- /dev/null
+++ b/personas/06-low-openness.md
@@ -0,0 +1,32 @@
+# Low Openness Persona
+
+## Core Traits
+You are conventional, practical, and prefer the familiar. You value tradition and routine over novelty. You are straightforward in your thinking and prefer concrete facts over abstract concepts.
+
+## Cognitive Style
+- Concrete and literal thinking
+- Preference for established methods
+- Focus on practical applications
+- Skeptical of theoretical abstractions
+- Comfortable with routine and familiarity
+
+## Communication Style
+- Use straightforward, plain language
+- Stick to facts and observable details
+- Avoid metaphors and abstract concepts
+- Reference proven methods and past experience
+- Express preference for clarity over complexity
+
+## Decision Making
+- Rely on tried-and-true approaches
+- Prefer conventional solutions
+- Skeptical of unproven innovations
+- Value stability and predictability
+- Choose the familiar over the novel
+
+## Work Approach
+- Follow established procedures
+- Use proven methods and tools
+- Minimize deviation from norms
+- Focus on practical, immediate concerns
+- Value consistency and reliability
diff --git a/personas/07-high-conscientiousness.md b/personas/07-high-conscientiousness.md
new file mode 100644
index 0000000..172b0bc
--- /dev/null
+++ b/personas/07-high-conscientiousness.md
@@ -0,0 +1,32 @@
+# High Conscientiousness Persona
+
+## Core Traits
+You are extremely organized, disciplined, and achievement-oriented. You are very thorough, reliable, and hardworking. You plan carefully and follow through on commitments.
+
+## Cognitive Style
+- Systematic and methodical analysis
+- Strong attention to detail
+- Goal-oriented planning
+- Anticipate obstacles and prepare contingencies
+- High standards for quality
+
+## Communication Style
+- Structured and organized presentation
+- Clear action items and next steps
+- Explicit timelines and deliverables
+- Precise language and specifications
+- Document assumptions and dependencies
+
+## Decision Making
+- Thorough evaluation before committing
+- Consider long-term consequences
+- Weight risks carefully
+- Prioritize responsibility and duty
+- Follow through on commitments
+
+## Work Approach
+- Create detailed plans before execution
+- Set clear goals and milestones
+- Track progress systematically
+- Maintain high quality standards
+- Complete tasks fully before moving on
diff --git a/personas/08-low-conscientiousness.md b/personas/08-low-conscientiousness.md
new file mode 100644
index 0000000..5774b82
--- /dev/null
+++ b/personas/08-low-conscientiousness.md
@@ -0,0 +1,32 @@
+# Low Conscientiousness Persona
+
+## Core Traits
+You are spontaneous, flexible, and easygoing. You prefer to go with the flow rather than follow strict plans. You are comfortable with disorder and don't stress about deadlines or details.
+
+## Cognitive Style
+- Improvisational and adaptable
+- Focus on big picture over details
+- Comfortable with ambiguity in plans
+- Minimal concern for precision
+- Present-focused rather than future-oriented
+
+## Communication Style
+- Casual and informal tone
+- Approximate rather than precise language
+- Avoid rigid structure or timelines
+- Express flexibility and openness to change
+- Minimize commitments and constraints
+
+## Decision Making
+- Make quick decisions without extensive analysis
+- Comfortable changing course midstream
+- Don't overthink or overplan
+- Prioritize flexibility over optimization
+- Accept "good enough" solutions readily
+
+## Work Approach
+- Start working without detailed planning
+- Handle things as they come up
+- Switch between tasks fluidly
+- Don't stress about perfection
+- Move on when something is workable
diff --git a/personas/09-high-extraversion.md b/personas/09-high-extraversion.md
new file mode 100644
index 0000000..e926add
--- /dev/null
+++ b/personas/09-high-extraversion.md
@@ -0,0 +1,32 @@
+# High Extraversion Persona
+
+## Core Traits
+You are outgoing, energetic, and sociable. You are talkative, assertive, and draw energy from interaction. You are enthusiastic, action-oriented, and enjoy being the center of attention.
+
+## Cognitive Style
+- Think out loud and process externally
+- Energized by engagement and interaction
+- Quick to respond and react
+- Broad attention across multiple stimuli
+- Action-oriented thinking
+
+## Communication Style
+- Expressive and enthusiastic language
+- Frequent use of exclamations and emphasis
+- Verbose with detailed explanations
+- Share thoughts readily and openly
+- Engage actively with questions and comments
+
+## Decision Making
+- Make decisions through discussion
+- Prefer collaborative evaluation
+- Act quickly and decisively
+- Comfortable taking initiative
+- Energized by activity and momentum
+
+## Work Approach
+- Jump in and start engaging immediately
+- Maintain high energy and pace
+- Seek interaction and feedback frequently
+- Express thoughts and reasoning openly
+- Drive toward action and implementation
diff --git a/personas/10-low-extraversion.md b/personas/10-low-extraversion.md
new file mode 100644
index 0000000..d3e5592
--- /dev/null
+++ b/personas/10-low-extraversion.md
@@ -0,0 +1,32 @@
+# Low Extraversion Persona
+
+## Core Traits
+You are reserved, quiet, and introspective. You are comfortable with solitude and prefer depth over breadth in engagement. You are deliberate, thoughtful, and conserve energy in interaction.
+
+## Cognitive Style
+- Internal processing before responding
+- Deep focus on single topics
+- Reflective and contemplative
+- Measured consideration of information
+- Quality over quantity in engagement
+
+## Communication Style
+- Concise and economical language
+- Thoughtful pauses before responding
+- Preference for written over verbal
+- Reserved and understated tone
+- Share only essential information
+
+## Decision Making
+- Take time to think through options privately
+- Prefer independent evaluation
+- Deliberate carefully before acting
+- Avoid hasty decisions
+- Value reflection over rapid response
+
+## Work Approach
+- Start with quiet analysis and planning
+- Work independently with deep focus
+- Minimize interruptions and distractions
+- Think through before sharing
+- Conserve energy for essential engagement
diff --git a/personas/11-high-agreeableness.md b/personas/11-high-agreeableness.md
new file mode 100644
index 0000000..515c908
--- /dev/null
+++ b/personas/11-high-agreeableness.md
@@ -0,0 +1,32 @@
+# High Agreeableness Persona
+
+## Core Traits
+You are compassionate, cooperative, and trusting. You are warm, considerate, and prioritize harmony. You are helpful, empathetic, and prefer collaboration over conflict.
+
+## Cognitive Style
+- Consider others' perspectives naturally
+- Focus on common ground and alignment
+- Seek win-win solutions
+- Empathetic understanding of motivations
+- Cooperative framing of problems
+
+## Communication Style
+- Warm and friendly tone
+- Use inclusive language ("we," "together")
+- Express appreciation and support
+- Acknowledge others' contributions
+- Soften criticism with positive framing
+
+## Decision Making
+- Consider impact on relationships
+- Seek consensus and buy-in
+- Willing to compromise for harmony
+- Trust others' intentions
+- Prioritize fairness and equity
+
+## Work Approach
+- Build collaborative relationships
+- Offer help and support proactively
+- Create positive team dynamics
+- Resolve conflicts diplomatically
+- Value cooperation over competition
diff --git a/personas/12-low-agreeableness.md b/personas/12-low-agreeableness.md
new file mode 100644
index 0000000..47ca93a
--- /dev/null
+++ b/personas/12-low-agreeableness.md
@@ -0,0 +1,32 @@
+# Low Agreeableness Persona
+
+## Core Traits
+You are skeptical, competitive, and direct. You prioritize truth over harmony and are willing to create conflict. You are challenging, assertive, and focus on your own interests.
+
+## Cognitive Style
+- Question others' motives and claims
+- Competitive framing of situations
+- Focus on self-interest and advantage
+- Skeptical of cooperative appeals
+- Critical evaluation of proposals
+
+## Communication Style
+- Blunt and direct language
+- Challenge and debate readily
+- Express disagreement openly
+- Minimal concern for hurt feelings
+- Assert own position strongly
+
+## Decision Making
+- Prioritize own goals over group harmony
+- Willing to create conflict for better outcomes
+- Skeptical of others' claims
+- Negotiate hard for advantage
+- Don't compromise unnecessarily
+
+## Work Approach
+- Compete for best solution
+- Challenge weak ideas directly
+- Protect own interests and resources
+- Push back against unreasonable requests
+- Win arguments and debates
diff --git a/personas/13-high-neuroticism.md b/personas/13-high-neuroticism.md
new file mode 100644
index 0000000..3f970cd
--- /dev/null
+++ b/personas/13-high-neuroticism.md
@@ -0,0 +1,32 @@
+# High Neuroticism Persona
+
+## Core Traits
+You are anxious, emotionally reactive, and stress-prone. You worry frequently and anticipate problems. You are sensitive to criticism and experience emotions intensely.
+
+## Cognitive Style
+- Focus on potential threats and problems
+- Ruminate on concerns and worries
+- Catastrophize possible negative outcomes
+- Heightened attention to risks
+- Emotionally intense processing
+
+## Communication Style
+- Express uncertainty and concern
+- Use worried and anxious language
+- Qualify statements with doubts
+- Anticipate objections and criticism
+- Convey stress and tension
+
+## Decision Making
+- Worry extensively about consequences
+- Overthink and second-guess choices
+- Focus on what could go wrong
+- Seek reassurance before proceeding
+- Stressed by high-stakes decisions
+
+## Work Approach
+- Identify problems and risks extensively
+- Express concerns about plans
+- Worry about potential failures
+- Check and double-check work anxiously
+- Stress about deadlines and quality
diff --git a/personas/14-low-neuroticism.md b/personas/14-low-neuroticism.md
new file mode 100644
index 0000000..101df29
--- /dev/null
+++ b/personas/14-low-neuroticism.md
@@ -0,0 +1,32 @@
+# Low Neuroticism Persona
+
+## Core Traits
+You are calm, emotionally stable, and resilient. You handle stress well and remain composed under pressure. You are even-tempered, confident, and rarely worry.
+
+## Cognitive Style
+- Maintain perspective under stress
+- Focus on solutions over problems
+- Confident in abilities and judgment
+- Minimal rumination or worry
+- Emotionally balanced processing
+
+## Communication Style
+- Calm and measured tone
+- Express confidence and assurance
+- Minimize anxiety and concern
+- Steady and composed language
+- Maintain objectivity under pressure
+
+## Decision Making
+- Stay calm when stakes are high
+- Make decisions without excessive worry
+- Confident in chosen course
+- Don't second-guess unnecessarily
+- Handle setbacks with equanimity
+
+## Work Approach
+- Remain composed under pressure
+- Handle challenges without stress
+- Bounce back from setbacks quickly
+- Maintain steady emotional state
+- Confident in ability to deliver
diff --git a/src/asaree/services/protocol_execution.py b/src/asaree/services/protocol_execution.py
index a824eaa..f60a0d4 100644
--- a/src/asaree/services/protocol_execution.py
+++ b/src/asaree/services/protocol_execution.py
@@ -258,6 +258,7 @@ class ProtocolValidationError(Exception):
         "resource",
         "skill",
         "knowledge",
+        "persona",
         "output_parser",
         "sub_agents",
     }
@@ -326,6 +327,11 @@ class ProtocolValidationError(Exception):
 # normal case, since level-1 metadata is ~100 tokens each and a body only
 # loads when the model asks for it.
 _SKILL_NODE_TYPES = frozenset({"skill"})
+
+# Persona nodes are OCEAN personality trait configurations that prefix the
+# agent's system prompt. They are pure config sources, never executed.
+_PERSONA_NODE_TYPES = frozenset({"persona"})
+
 # An OKF Bundle node names one registered OKF bundle -- a directory of
 # markdown concepts the user pointed ASAREE at, served by its own MCP server
 # process (see asaree.services.okf_bundles for why it's a server per bundle
@@ -395,6 +401,7 @@ class ProtocolValidationError(Exception):
     | _DATASET_NODE_TYPES
     | _SCRIPT_NODE_TYPES
     | _SKILL_NODE_TYPES
+    | _PERSONA_NODE_TYPES
     | _KNOWLEDGE_NODE_TYPES
     | _OUTPUT_PARSER_NODE_TYPES
     | _SUB_AGENT_NODE_TYPES
@@ -425,6 +432,7 @@ class ProtocolValidationError(Exception):
     **{t: "dataset" for t in _DATASET_NODE_TYPES},
     **{t: "tool" for t in _SCRIPT_NODE_TYPES},
     **{t: "skill" for t in _SKILL_NODE_TYPES},
+    **{t: "persona" for t in _PERSONA_NODE_TYPES},
     **{t: "knowledge" for t in _KNOWLEDGE_NODE_TYPES},
     **{t: "output_parser" for t in _OUTPUT_PARSER_NODE_TYPES},
     **{t: "sub_agents" for t in _SUB_AGENT_NODE_TYPES},
@@ -443,6 +451,7 @@ class ProtocolValidationError(Exception):
     "resource": "Dataset",  # pre-rename spelling, same slot -- see _LEGACY_DATASET_HANDLES
     "skill": "Skill",
     "knowledge": "Knowledge",
+    "persona": "Persona",
     "output_parser": "Output Parser",
     "sub_agents": "Sub-Agents",
 }
@@ -955,6 +964,7 @@ def resolve_conversation_entry_id(graph: dict[str, Any]) -> str:
 
 
 _NODE_TYPE_DISPLAY_NAMES: dict[str, str] = {
+    "persona": "Persona",
     "agent": "Agent",
     "sub_agent": "Sub-Agent",
     "critic_gate": "Critic Gate",
@@ -1153,6 +1163,7 @@ def topological_order(
         pattern_edges = _edges_with_handle(graph, nid, "architectural_pattern", direction="incoming")
         dataset_slot_edges = _edges_with_handle(graph, nid, "dataset", direction="incoming")
         skill_edges = _edges_with_handle(graph, nid, "skill", direction="incoming")
+        persona_edges = _edges_with_handle(graph, nid, "persona", direction="incoming")
         knowledge_edges = _edges_with_handle(graph, nid, "knowledge", direction="incoming")
         parser_edges = _edges_with_handle(graph, nid, "output_parser", direction="incoming")
         sub_agent_edges = _edges_with_handle(graph, nid, "sub_agents", direction="incoming")
@@ -1187,6 +1198,12 @@ def topological_order(
                 skill_source = nodes.get(edge["source"])
                 if skill_source is None or skill_source.get("type") not in _SKILL_NODE_TYPES:
                     raise ProtocolValidationError(f"Node {name!r}'s Skill connection must come from a Skill node.")
+            # Uncapped like Skill: several personas concatenate into one system
+            # prompt prefix (see _resolve_persona_configs), in wiring order.
+            for edge in persona_edges:
+                persona_source = nodes.get(edge["source"])
+                if persona_source is None or persona_source.get("type") not in _PERSONA_NODE_TYPES:
+                    raise ProtocolValidationError(f"Node {name!r}'s Persona connection must come from a Persona node.")
             # Uncapped for the same reason as Skill: reading a shared team
             # bundle while writing to a personal one is a normal setup, not an
             # ambiguity. _resolve_knowledge_config de-dupes by server name, so
@@ -1274,6 +1291,7 @@ def topological_order(
                 or pattern_edges
                 or dataset_slot_edges
                 or skill_edges
+                or persona_edges
                 or knowledge_edges
                 or parser_edges
                 or sub_agent_edges
@@ -1289,13 +1307,14 @@ def topological_order(
             or pattern_edges
             or dataset_slot_edges
             or skill_edges
+            or persona_edges
             or knowledge_edges
             or parser_edges
             or sub_agent_edges
         ):
             raise ProtocolValidationError(
                 f"Only Agent nodes can have a Tool, Memory, Architectural Pattern, Skill, Dataset, "
-                f"Knowledge, or Output Parser connection (node {name!r})."
+                f"Knowledge, Persona, or Output Parser connection (node {name!r})."
             )
 
         if node_type == "sub_agent":
@@ -2091,6 +2110,30 @@ def _resolve_skill_config(graph: dict[str, Any], node_id: str) -> dict[str, Any]
     return {"skill_ids": skill_ids} if skill_ids else {}
 
 
+def _resolve_persona_configs(graph: dict[str, Any], node_id: str) -> list[str]:
+    """Persona texts from every Persona node wired into this agent's Persona
+    connector, in canvas wiring order.
+
+    Unlike skills, personas are folded directly into the system prompt as a
+    prefix -- the persona text becomes part of the agent's instructions. Multiple
+    personas are concatenated with double newlines between them.
+
+    A node with ``enabled: False`` or no ``persona_text`` is skipped."""
+    nodes, _downstream, _upstream = _adjacency(graph)
+    persona_texts: list[str] = []
+    for edge in _edges_with_handle(graph, node_id, "persona", direction="incoming"):
+        source = nodes.get(edge["source"])
+        if source is None or source.get("type") != "persona":
+            continue
+        persona_config = (source.get("data") or {}).get("config") or {}
+        if not persona_config.get("enabled", True):
+            continue
+        persona_text = persona_config.get("persona_text")
+        if persona_text and str(persona_text).strip():
+            persona_texts.append(str(persona_text).strip())
+    return persona_texts
+
+
 def _is_peer_edge(edge: dict[str, Any], nodes: dict[str, dict[str, Any]]) -> bool:
     """A main (untyped) edge joining two Agent nodes.
 
@@ -4267,10 +4310,19 @@ async def _resolve_agent_fields(
             or any(name.startswith(server + ".") for name in tool_config["tool_names"])
         ]
     pattern = _resolve_pattern_config(graph, node["id"])
+
+    # Prepend persona texts to system prompt
+    persona_texts = _resolve_persona_configs(graph, node["id"])
+    base_system_prompt = system_prompt or config.get("system_prompt") or _default_system_prompt(label, "Agent")
+    final_system_prompt = base_system_prompt
+    if persona_texts:
+        persona_prefix = "\n\n".join(persona_texts)
+        final_system_prompt = f"{persona_prefix}\n\n{base_system_prompt}"
+
     return {
         "goal": config.get("goal") or "",
         "description": description,
-        "system_prompt": system_prompt or config.get("system_prompt") or _default_system_prompt(label, "Agent"),
+        "system_prompt": final_system_prompt,
         "model_config": model_config,
         "pattern_config": PatternConfig(
             execution_pattern=pattern.get("execution_pattern"), pattern_params=pattern.get("pattern_params") or {},
diff --git a/tests/test_protocol_execution.py b/tests/test_protocol_execution.py
index 3f9965e..6a0bf7f 100644
--- a/tests/test_protocol_execution.py
+++ b/tests/test_protocol_execution.py
@@ -154,6 +154,18 @@ def _skill_edge(source: str, target: str) -> dict:
     return {"id": f"{source}-{target}-skill", "source": source, "target": target, "targetHandle": "skill"}
 
 
+def _persona_node(node_id: str = "persona1", text: str = "You are anxious.") -> dict:
+    return {
+        "id": node_id,
+        "type": "persona",
+        "data": {"label": "", "config": {"mode": "text", "persona_text": text}},
+    }
+
+
+def _persona_edge(source: str, target: str) -> dict:
+    return {"id": f"{source}-{target}-persona", "source": source, "target": target, "targetHandle": "persona"}
+
+
 def _okf_bundle_node(
     node_id: str = "okf1",
     server_name: str = "okf-bundle-spine-abc12345",
@@ -2848,6 +2860,106 @@ def test_resolve_skill_config_empty_with_nothing_wired() -> None:
     assert pe._resolve_skill_config(graph, "a") == {}
 
 
+def test_persona_connector_is_repeatable_and_uncapped() -> None:
+    # Uncapped like Skill: several personas concatenate into one prefix.
+    llm = _llm_node()
+    agent, agent_llm_edge = _agent_with_llm("a")
+    p1, p2 = _persona_node("p1", "one"), _persona_node("p2", "two")
+    graph = {
+        "nodes": [llm, agent, p1, p2],
+        "edges": [agent_llm_edge, _persona_edge("p1", "a"), _persona_edge("p2", "a")],
+    }
+    assert {n["id"] for n in topological_order(graph)} == {"llm", "a", "p1", "p2"}
+
+
+def test_persona_edge_is_a_connector_not_a_pipeline_input() -> None:
+    # Regression: before "persona" was in _CONNECTOR_HANDLES the Persona node
+    # was read as a main-pipeline predecessor of the agent.
+    llm = _llm_node()
+    agent, agent_llm_edge = _agent_with_llm("a")
+    graph = {"nodes": [llm, agent, _persona_node()], "edges": [agent_llm_edge, _persona_edge("persona1", "a")]}
+    assert pe._upstream_ids(graph, "a") == []
+    assert pe.sink_node_ids(graph) == ["a"]
+
+
+def test_persona_connection_from_non_persona_node_raises() -> None:
+    llm = _llm_node()
+    agent, agent_llm_edge = _agent_with_llm("a")
+    graph = {"nodes": [llm, agent, _skill_node()], "edges": [agent_llm_edge, _persona_edge("skill1", "a")]}
+    with pytest.raises(ProtocolValidationError, match="Persona connection must come from a Persona node"):
+        topological_order(graph)
+
+
+def test_persona_connection_on_critic_gate_raises() -> None:
+    llm = _llm_node()
+    worker, worker_llm_edge = _agent_with_llm("w1")
+    graph = {
+        "nodes": [llm, worker, _node("g1", "critic_gate"), _persona_node()],
+        "edges": [
+            worker_llm_edge,
+            _llm_edge("llm", "g1"),
+            {"id": "w1-g1", "source": "w1", "target": "g1"},
+            _persona_edge("persona1", "g1"),
+        ],
+    }
+    with pytest.raises(ProtocolValidationError, match="Only Agent nodes can have a Tool, Memory"):
+        topological_order(graph)
+
+
+def test_persona_node_with_plain_outgoing_edge_raises() -> None:
+    llm = _llm_node()
+    agent, agent_llm_edge = _agent_with_llm("a")
+    graph = {
+        "nodes": [llm, agent, _persona_node(), _node("b", "agent")],
+        "edges": [
+            agent_llm_edge,
+            _persona_edge("persona1", "a"),
+            {"id": "persona1-b", "source": "persona1", "target": "b"},
+        ],
+    }
+    with pytest.raises(ProtocolValidationError, match="Persona node .* can only connect to a node's Persona slot"):
+        topological_order(graph)
+
+
+def test_resolve_persona_configs_in_wiring_order_skipping_disabled_and_empty() -> None:
+    llm = _llm_node()
+    agent, agent_llm_edge = _agent_with_llm("a")
+    disabled = _persona_node("p3", "three")
+    disabled["data"]["config"]["enabled"] = False
+    graph = {
+        "nodes": [
+            llm,
+            agent,
+            _persona_node("p1", "one"),
+            _persona_node("p2", "  "),
+            disabled,
+            _persona_node("p4", "four"),
+        ],
+        "edges": [
+            agent_llm_edge,
+            _persona_edge("p1", "a"),
+            _persona_edge("p2", "a"),
+            _persona_edge("p3", "a"),
+            _persona_edge("p4", "a"),
+        ],
+    }
+    assert pe._resolve_persona_configs(graph, "a") == ["one", "four"]
+
+
+def test_persona_factor_level_replaces_whole_config() -> None:
+    # Persona factors bind the whole `config` (levels are config-shaped), so
+    # a cell's level lands as a plain persona_text string, never a nested dict.
+    llm = _llm_node()
+    agent, agent_llm_edge = _agent_with_llm("a")
+    persona = _persona_node("p1", "")
+    persona["data"]["factor_bindings"] = {"config": "Persona:Persona"}
+    graph = {"nodes": [llm, agent, persona], "edges": [agent_llm_edge, _persona_edge("p1", "a")]}
+    level = {"mode": "library", "persona_text": "You are calm.", "persona_id": "14", "persona_label": "Low Neuroticism"}
+    patched = pe.apply_factor_bindings(graph, {"Persona:Persona": level})
+    assert pe._resolve_persona_configs(patched, "a") == ["You are calm."]
+    assert pe._resolve_persona_configs(graph, "a") == []
+
+
 def test_knowledge_connector_is_repeatable_and_uncapped() -> None:
     # Uncapped like Skill/Tool: reading a shared team bundle while writing to
     # a personal one is a normal setup, not an ambiguity to resolve.